Legal claims defining the scope of protection. Each claim is shown in both the original legal language and a plain English translation.
1. A system for machine-learning-based atrial fibrillation detection with the aid of a digital computer, comprising: a database operable to maintain a plurality of electrocardiography (ECG) features and annotated patterns of the features, at least some of the patterns associated with atrial fibrillation; at least one server interconnected to the database, the at least one server configured to: train a classifier based on the annotated patterns in the database; receive a representation of an ECG signal recorded by an ambulatory monitor recorder during a plurality of temporal windows; detect a plurality of the ECG features in at least some of the portions of the representation falling within each of the temporal windows; use the trained classifier to identify patterns of the ECG features within one or more of the portions of the ECG signal; for each of the portions, calculate a value indicative of whether the portion of the representation within that ECG signal is associated the patient experiencing atrial fibrillation; calculate a further value indicative of whether the portion of the representation within that ECG signal is associated with the patient not experiencing atrial fibrillation; compare the further value to the value; determine that the portion of the ECG signal is associated with the patient experiencing atrial fibrillation based on the comparison; and take an action based on the determination that the portion of the ECG signal is associated with the patient experiencing atrial fibrillation.
2. A system for machine-learning-based atrial fibrillation detection according to claim 1 , the at least one server further configured to: obtain training data comprising a plurality of the ECG features and a plurality of patterns of the ECG features; and obtain annotations of patterns of the ECG features in the training data, wherein the training of the classifier is based on the annotations.
This system relates to machine-learning-based atrial fibrillation (AF) detection using electrocardiogram (ECG) data. AF is a common heart condition characterized by irregular heart rhythms, and accurate detection is critical for diagnosis and treatment. The system addresses the challenge of reliably identifying AF patterns in ECG signals, which can be complex and variable. The system includes at least one server configured to process ECG data for AF detection. The server obtains training data comprising multiple ECG features and patterns derived from these features. These features may include heart rate variability, P-wave morphology, and other ECG signal characteristics. The system also collects annotations of these patterns, which are used to train a classifier. The annotations label the patterns as indicative of AF or normal sinus rhythm, enabling the classifier to learn distinguishing features. During operation, the system extracts ECG features from input signals and applies the trained classifier to detect AF. The classifier is optimized through supervised learning, where the annotated training data improves its accuracy in distinguishing AF patterns from non-AF patterns. This approach enhances the reliability of AF detection, supporting early diagnosis and intervention. The system may integrate with wearable or clinical ECG devices to provide real-time or batch analysis of ECG data.
3. A system for machine-learning-based atrial fibrillation detection according to claim 1 , the at least one server further configured to: test an accuracy of the trained classifier and perform further training based on a result of the test.
A system for machine-learning-based atrial fibrillation detection analyzes physiological data to identify atrial fibrillation (AF) episodes. The system includes at least one server that processes input data, such as electrocardiogram (ECG) signals, to train a machine-learning classifier. The classifier is trained using labeled data to distinguish between normal sinus rhythm and AF patterns. The system further evaluates the trained classifier's accuracy by testing it on validation data. If the accuracy does not meet a predefined threshold, the system performs additional training iterations to refine the classifier. This iterative process improves the model's performance in detecting AF episodes. The system may also incorporate additional data sources, such as patient demographics or historical medical records, to enhance detection accuracy. The goal is to provide a reliable, automated tool for early AF detection, reducing the need for manual analysis and improving diagnostic efficiency in clinical settings.
4. A system for machine-learning-based atrial fibrillation detection according to claim 1 , wherein the determination is made upon the value exceeding the further value.
Atrial fibrillation (AF) is a common cardiac arrhythmia that requires accurate and timely detection for effective treatment. Traditional methods of AF detection often rely on electrocardiogram (ECG) analysis, but these can be time-consuming and require expert interpretation. Machine learning (ML) offers a promising alternative by automating AF detection from ECG data, but existing systems may lack robustness in distinguishing AF from other arrhythmias or noise. This system improves upon prior art by using machine learning to analyze ECG signals and detect atrial fibrillation with enhanced accuracy. The system processes raw ECG data, extracts relevant features, and applies a trained ML model to classify the signal as AF or non-AF. A key innovation is the use of a threshold-based decision mechanism, where the system compares a computed value (e.g., a confidence score or feature metric) against a predefined threshold. If the value exceeds this threshold, the system confirms AF detection, reducing false positives and improving reliability. The system may also incorporate additional validation steps, such as comparing the computed value against a secondary threshold or cross-referencing with other physiological signals, to further refine detection accuracy. This approach ensures that AF is identified with higher confidence while minimizing misclassification errors. The system is designed to operate in real-time or near-real-time, making it suitable for wearable or clinical monitoring devices.
5. A system for machine-learning-based atrial fibrillation detection according to claim 1 , wherein the action comprises sending an alert regarding the determination.
This invention relates to a machine-learning-based system for detecting atrial fibrillation (AF), a common heart rhythm disorder. The system addresses the challenge of accurately and timely identifying AF episodes, which can be difficult to detect using traditional methods due to their intermittent nature and subtle symptoms. The system leverages machine learning algorithms to analyze physiological data, such as electrocardiogram (ECG) signals, to detect AF with high accuracy. The core functionality involves processing input data through a trained machine learning model to determine whether AF is present. If AF is detected, the system generates an action, such as sending an alert to a user or healthcare provider. This alert can be transmitted via a communication interface to a mobile device, medical monitoring system, or other output device, ensuring timely intervention. The system may also include preprocessing steps to enhance data quality and post-processing steps to refine detection results. By automating AF detection and alerting, the system improves patient monitoring and reduces the risk of complications associated with undiagnosed or untreated AF. The invention is particularly useful in wearable or remote monitoring devices, where continuous or periodic AF detection is critical.
6. A system for machine-learning-based atrial fibrillation detection according to claim 1 , the at least one server further configured to: generate a matrix with the identified features and the patterns; and generate at least one matrix with weights for the identified features and patterns, wherein the value and the further value are calculated using the weight matrix.
This system detects atrial fibrillation (AF) using machine learning. AF is a heart rhythm disorder that can lead to serious complications if undiagnosed. The system addresses the challenge of accurately identifying AF from physiological signals, such as electrocardiograms (ECGs), by leveraging machine learning to analyze patterns in the data. The system includes at least one server that processes physiological signals to extract relevant features and patterns indicative of AF. These features and patterns are organized into a matrix. The server then generates a weight matrix, where each weight corresponds to the importance of a specific feature or pattern in detecting AF. The system calculates values and further values (e.g., probabilities or scores) using this weight matrix to determine the likelihood of AF presence. The machine learning model is trained to assign appropriate weights to features and patterns, improving detection accuracy. The system may also incorporate additional data, such as patient history or signal quality metrics, to enhance reliability. By automating AF detection, the system reduces the need for manual analysis, enabling faster and more consistent diagnosis. The approach is particularly useful in clinical settings where timely AF detection is critical for patient management.
7. A system for machine-learning-based atrial fibrillation detection according to claim 1 , wherein each of the temporal windows is between 2 and 60 seconds.
Atrial fibrillation (AF) is a common cardiac arrhythmia characterized by irregular heart rhythms, which can lead to serious health complications if undetected. Traditional AF detection methods often rely on electrocardiogram (ECG) analysis, but these approaches may lack real-time monitoring capabilities or require specialized equipment. Machine learning (ML) offers a promising alternative for automated AF detection, but existing systems may struggle with optimal temporal window selection for accurate classification. This invention describes a machine-learning-based system for detecting atrial fibrillation in cardiac signals, such as ECG data. The system processes the input signal by dividing it into multiple temporal windows, each ranging between 2 and 60 seconds in duration. Within each window, the system extracts relevant features from the cardiac signal, which are then fed into a trained machine learning model. The model analyzes these features to determine whether atrial fibrillation is present. The system may also include preprocessing steps to enhance signal quality, such as noise reduction or artifact removal, before feature extraction. The use of adjustable temporal windows allows the system to balance between computational efficiency and detection accuracy, ensuring reliable AF detection in real-time or near-real-time applications. This approach improves upon prior methods by optimizing the temporal resolution for machine learning-based classification, making it suitable for wearable or remote monitoring devices.
8. A system for machine-learning-based atrial fibrillation detection according to claim 1 , wherein the database comprises 32 of the ECG features.
Atrial fibrillation (AF) is a common cardiac arrhythmia that requires accurate detection for effective diagnosis and treatment. Traditional methods of AF detection rely on manual analysis of electrocardiogram (ECG) signals, which can be time-consuming and prone to human error. Machine learning (ML) techniques offer a more efficient and scalable approach to AF detection by analyzing ECG features extracted from patient data. The system uses a machine-learning model trained on a database of ECG features to detect atrial fibrillation. The database includes 32 distinct ECG features, which are derived from raw ECG signals through signal processing techniques. These features capture various characteristics of the ECG waveform, such as heart rate variability, P-wave morphology, and R-R interval irregularities, which are indicative of AF. The machine-learning model processes these features to classify whether a patient's ECG signal exhibits signs of atrial fibrillation. The system may also include preprocessing steps to enhance the quality of the ECG signals before feature extraction, such as noise reduction and baseline correction. Additionally, the model may be trained using supervised learning techniques, where labeled ECG data from patients with and without AF is used to optimize the model's performance. The output of the system is a binary classification indicating the presence or absence of AF, which can be used by healthcare professionals to assist in diagnosis. This approach improves the accuracy and efficiency of AF detection compared to traditional methods, enabling earlier intervention and better patient outcomes. The use of 32 specific ECG features ensures that the model captures a comprehensive set of relevant information for reliable AF detection.
9. A system for machine-learning-based atrial fibrillation detection according to claim 1 , the at least one server further configured to: perform a noise filtering of at least some of the portions of the ECG signal prior to identification of the ECG features.
A system for detecting atrial fibrillation (AF) using machine learning processes electrocardiogram (ECG) signals to identify AF episodes. The system includes at least one server that receives ECG data from one or more wearable or implantable medical devices. The server processes the ECG signals to extract relevant features, such as heart rate variability, P-wave morphology, and irregular RR intervals, which are indicative of AF. A trained machine learning model analyzes these features to classify the ECG signal as AF or non-AF. The system may also incorporate additional data, such as patient demographics or historical ECG records, to improve detection accuracy. To enhance reliability, the server performs noise filtering on the ECG signal before feature extraction, removing artifacts caused by motion, electrical interference, or other disturbances. The filtered signal is then analyzed to identify key ECG features used for AF detection. The system may further include a user interface to display detection results, alerts, or recommendations for further medical evaluation. This approach enables continuous, automated AF monitoring, reducing the need for manual interpretation and improving early detection of AF episodes.
10. A method for machine-learning-based atrial fibrillation detection with the aid of a digital computer, comprising: maintaining in a database a plurality of electrocardiography (ECG) features and annotated patterns of the features, at least some of the patterns associated with atrial fibrillation; training by an at least one server connected to the database a classifier based on the annotated patterns in the database; receiving by the at least one server a representation of an ECG signal recorded by an ambulatory monitor recorder during a plurality of temporal windows; detecting by the at least one server a plurality of the ECG features in at least some of the portions of the representation falling within each of the temporal windows; using by the at least one server the trained classifier to identify patterns of the ECG features within one or more of the portions of the ECG signal; for each of the portions, calculating by the at least one server a value indicative of whether the portion of the representation within that ECG signal is associated the patient experiencing atrial fibrillation; calculating by the at least one server a further value indicative of whether the portion of the representation within that ECG signal is associated with the patient not experiencing atrial fibrillation; comparing the further value to the score; determining that the portion of the ECG signal is associated with the patient experiencing atrial fibrillation based on the comparison; taking by the at least one server an action based on the determination that the portion of the ECG signal is associated with the patient experiencing atrial fibrillation.
Atrial fibrillation (AF) is a common cardiac arrhythmia that requires accurate and timely detection for effective management. Traditional methods of AF detection often rely on manual analysis of electrocardiography (ECG) signals, which can be time-consuming and prone to human error. Machine learning offers a potential solution by automating the detection process, but existing systems may lack robustness in handling real-world ECG data, particularly from ambulatory monitors that record signals over extended periods. This invention describes a machine-learning-based system for detecting atrial fibrillation using ECG signals recorded by ambulatory monitors. The system maintains a database of ECG features and annotated patterns, where some patterns are associated with AF. A classifier is trained on this database to recognize AF-related patterns. When an ECG signal is received, the system processes it in multiple temporal windows, extracting features from each window. The trained classifier then identifies patterns within these features. For each window, the system calculates two values: one indicating the likelihood of AF and another indicating the likelihood of non-AF. These values are compared to determine whether AF is present. If AF is detected, the system takes an appropriate action, such as alerting healthcare providers or adjusting treatment protocols. This approach improves the accuracy and efficiency of AF detection in ambulatory settings, enabling earlier intervention and better patient outcomes.
11. A method for machine-learning-based atrial fibrillation detection according to claim 10 , further comprising: obtaining training data comprising a plurality of the ECG features and a plurality of patterns of the ECG features; and obtaining annotations of patterns of the ECG features in the training data, wherein the training of the classifier is based on the annotations.
This invention relates to machine-learning-based atrial fibrillation (AF) detection using electrocardiogram (ECG) data. AF is a common heart condition characterized by irregular and often rapid heartbeats, which can be challenging to detect accurately. The invention addresses the need for improved AF detection by leveraging machine learning to analyze ECG features and patterns. The method involves obtaining training data that includes multiple ECG features and patterns derived from ECG signals. These features may include time-domain, frequency-domain, or morphological characteristics of the ECG waveform. The training data is annotated to label patterns associated with AF, providing ground truth for supervised learning. A classifier is trained using this annotated data to distinguish between normal and AF-affected ECG patterns. The trained classifier can then be applied to new ECG data to detect AF with high accuracy. The method may also involve preprocessing ECG signals to extract relevant features, such as heart rate variability, P-wave morphology, or irregular RR intervals, which are indicative of AF. The classifier may be a neural network, support vector machine, or another machine-learning model optimized for ECG analysis. By learning from annotated patterns, the system improves detection performance compared to traditional rule-based approaches. This technique enhances early AF diagnosis, enabling timely medical intervention.
12. A method for machine-learning-based atrial fibrillation detection according to claim 10 , further comprising: test an accuracy of the trained classifier and performing further training based on a result of the test.
Atrial fibrillation (AF) is a common cardiac arrhythmia that requires accurate detection for effective diagnosis and treatment. Traditional methods of AF detection often rely on manual analysis of electrocardiogram (ECG) data, which can be time-consuming and prone to human error. Machine learning (ML) techniques offer a promising solution by automating the detection process, but ensuring the accuracy of these models remains a challenge. This invention describes a method for improving the accuracy of machine-learning-based atrial fibrillation detection. The method involves training a classifier using labeled ECG data, where the labels indicate the presence or absence of AF. The classifier is trained to distinguish between normal sinus rhythm and AF patterns in the ECG signals. Once trained, the classifier's accuracy is tested using a separate validation dataset. If the accuracy does not meet a predefined threshold, the classifier undergoes further training to refine its performance. This iterative process ensures that the model continues to improve until it achieves a satisfactory level of accuracy in detecting AF. The method leverages supervised learning techniques, where the classifier is trained on a dataset with known outcomes. The testing phase evaluates the classifier's performance, and the results guide additional training to enhance detection accuracy. This approach helps mitigate false positives and false negatives, which are critical in medical diagnostics. By continuously refining the model, the method ensures reliable and consistent AF detection, supporting early intervention and better patient outcomes.
13. A method for machine-learning-based atrial fibrillation detection according to claim 10 , wherein the determination is made upon the value exceeding the further value.
Atrial fibrillation (AF) is a common cardiac arrhythmia characterized by irregular and often rapid heartbeats, which can lead to serious complications if undetected. Traditional AF detection methods rely on electrocardiogram (ECG) analysis, but these approaches may lack real-time accuracy or require specialized equipment. Machine learning (ML) offers a promising alternative by leveraging data-driven models to improve detection efficiency and reliability. This invention describes a machine-learning-based method for detecting atrial fibrillation. The method involves processing physiological signals, such as ECG or photoplethysmogram (PPG) data, to extract relevant features. A trained ML model analyzes these features to determine the presence of AF. The model generates a value representing the likelihood of AF, which is then compared to a predefined threshold. If the value exceeds this threshold, the system confirms AF detection. Additionally, the method may incorporate a secondary threshold or further value to refine the decision, ensuring higher accuracy by reducing false positives or negatives. The approach may also include preprocessing steps, such as noise reduction or signal normalization, to enhance data quality before analysis. By integrating ML with physiological signal processing, this method provides an automated, scalable solution for early AF detection, potentially improving patient outcomes through timely intervention.
14. A method for machine-learning-based atrial fibrillation detection according to claim 10 , wherein the action comprises sending an alert of the regarding the determination.
This technical summary describes a machine-learning-based system for detecting atrial fibrillation (AF), a common heart rhythm disorder. The system addresses the challenge of accurately identifying AF episodes in real-time, which is critical for timely medical intervention. The method involves analyzing physiological signals, such as electrocardiogram (ECG) data, to detect patterns indicative of AF. A machine-learning model processes these signals to classify them as AF or non-AF, using features extracted from the raw data. The model is trained on labeled datasets to improve accuracy and reliability. Once AF is detected, the system triggers an action, such as sending an alert to a healthcare provider or the patient. The alert includes details about the detection, such as the time, severity, and confidence level of the diagnosis. This ensures that appropriate medical responses can be initiated promptly. The system may also log the detection for further analysis or integrate with wearable devices to provide continuous monitoring. The method leverages advanced machine-learning techniques, including deep learning or traditional classifiers, to enhance detection accuracy. It may incorporate signal preprocessing steps to reduce noise and improve feature extraction. The system is designed to operate efficiently on edge devices or cloud platforms, depending on the application requirements. By automating AF detection, the system reduces the burden on healthcare professionals and improves patient outcomes through early intervention.
15. A method for machine-learning-based atrial fibrillation detection according to claim 10 , further comprising: generating a matrix with the identified features and the patterns; and generating at least one matrix with weights for the identified features and patterns, wherein the value and the further value are calculated using the weight matrix.
This invention relates to machine-learning-based atrial fibrillation (AF) detection, addressing the challenge of accurately identifying AF episodes from physiological signals. The method involves processing input data, such as electrocardiogram (ECG) or photoplethysmogram (PPG) signals, to extract relevant features and patterns indicative of AF. These features and patterns are then used to generate a matrix representing their relationships. Additionally, a weight matrix is created to assign importance to the identified features and patterns, where the weight values are used to calculate output values for AF detection. The method leverages machine learning to analyze the input data, extract meaningful features, and apply learned weights to improve detection accuracy. The weight matrix ensures that the most relevant features and patterns contribute more significantly to the final AF detection result, enhancing the reliability of the diagnostic process. This approach aims to provide a robust and automated solution for AF detection, reducing the need for manual interpretation and improving diagnostic efficiency.
16. A method for machine-learning-based atrial fibrillation detection according to claim 10 , wherein each of the temporal windows is between 2 and 60 seconds.
This invention relates to a machine-learning-based method for detecting atrial fibrillation (AF), a common heart rhythm disorder. The method addresses the challenge of accurately identifying AF episodes from physiological signals, such as electrocardiogram (ECG) data, in real-time or near-real-time applications. Traditional AF detection methods often struggle with false positives or require extensive computational resources, making them impractical for wearable or portable devices. The method processes physiological signals by dividing them into temporal windows, each lasting between 2 and 60 seconds. Within each window, the signal is analyzed using a machine-learning model trained to distinguish AF patterns from normal sinus rhythm or other arrhythmias. The model may incorporate features such as heart rate variability, signal morphology, or frequency-domain characteristics to improve detection accuracy. The temporal window size is optimized to balance computational efficiency and detection sensitivity, ensuring reliable AF identification without excessive processing overhead. The approach is designed for integration into medical devices, wearables, or remote monitoring systems, enabling early AF detection and intervention. By leveraging machine learning, the method adapts to individual patient variations and improves over time with additional data. The invention aims to enhance diagnostic accuracy while reducing the burden on healthcare systems by automating AF screening.
17. A method for machine-learning-based atrial fibrillation detection according to claim 10 , wherein the database comprises 32 of the ECG features.
Atrial fibrillation (AF) is a common cardiac arrhythmia that requires accurate detection for effective treatment. Traditional methods of AF detection rely on manual analysis of electrocardiogram (ECG) signals, which can be time-consuming and prone to human error. Machine learning (ML) techniques offer a more efficient and scalable approach to AF detection by analyzing ECG features automatically. However, the effectiveness of ML models depends on the quality and relevance of the input features. This invention describes a method for improving AF detection using machine learning by optimizing the set of ECG features used as input to the model. The method involves selecting a specific subset of 32 ECG features from a larger pool of possible features. These features are derived from raw ECG signals and may include time-domain, frequency-domain, and nonlinear characteristics that are known to be relevant to AF detection. The selected features are then used to train a machine learning model, such as a classifier, to distinguish between AF and non-AF ECG signals. The model is trained using a database of labeled ECG recordings, where each recording is annotated as either AF or non-AF. The trained model can then be applied to new, unseen ECG signals to predict the presence or absence of AF with improved accuracy compared to methods using a broader or less optimized set of features. By focusing on a carefully chosen subset of 32 ECG features, this method enhances the efficiency and performance of AF detection, reducing false positives and negatives while maintaining computational efficiency. This approach is particularly useful in clinical settings where rapid and reliable AF detection is critical for patient care.
18. A method for machine-learning-based atrial fibrillation detection according to claim 10 , further comprising: performing a noise filtering of at least some of the portions of the ECG signal prior to identification of the ECG features.
This invention relates to machine-learning-based atrial fibrillation (AF) detection using electrocardiogram (ECG) signals. AF is a common heart condition characterized by irregular and often rapid heart rhythms, which can be challenging to detect accurately. The invention addresses the need for improved AF detection by leveraging machine learning techniques to analyze ECG signals, while also mitigating the impact of noise that can distort the signal and reduce detection accuracy. The method involves processing an ECG signal to identify features indicative of AF. Before analyzing the signal, the method includes a noise filtering step to enhance the quality of the ECG data. This filtering step removes or reduces unwanted noise from the signal, ensuring that the subsequent feature extraction and machine learning analysis are based on cleaner, more reliable data. The filtered ECG signal is then divided into portions, and specific features are extracted from these portions to train or evaluate a machine learning model. The model is designed to classify the ECG signal as either AF or non-AF based on the extracted features. By incorporating noise filtering, the method improves the robustness and accuracy of AF detection, making it more reliable for clinical applications.
Unknown
November 5, 2019
Browse 5M+ US patents with plain-English claim translations and AI-generated analysis.